AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[Jay139]

Interesting
I found this thread via YT and decided to subscribe.
Wil there be a there a Ryzen APU also ?

 

[Crumpet]

There is no linear relation to the number of utilized cores (i.e core dependent boost bins like on Intel). Just MSCF (maximum single core frequency) and MACF (maximum all core frequency).

My brain has the consistency of extremely overboiled cabbage today..Please humour me.

So you're saying that if you Purchase a Ryzen cpu listed as 3.6/4.0ghz, then that is 4.0 MACF?

 

[Jan Olšan]

4.0 GHz should be maximum single core frequency, maximum all core frequency is probably not going to be advertised (that's how it is on FX/APUs).

 

[CatMerc]

Based on comparing apples to oranges? Sigh, you really like these comparisons.

Could at least post source: http://tieba.baidu.com/p/4975965039

Anyways, if all of these are unlocked, i really fail to understand who would buy higher bins.

I'm wondering if XFR offers something manual OC can't really properly provide, at least not with any reasonable convinience.
For example, if all cores can turbo up to 4.5, but a single core can do 4.8. You can always go into the BIOS and disable the rest of the cores to reach that 4.8, and then switch back to 4.5 with all cores when you're done with the task, but that would be incredibly annoying for everyday use.

If XFR can detect load on a specific core and boost it to the edge, while powering down the rest of the cores to a level that allows that single core to reach 4.8GHz, then it does something that would be straight up annoying to attempt in manual overclocking.

Basically, if XFR can optimize itself for specific application loads, then I'm getting a SKU with one. As fun as overclocking is, I'd like for my computer to be able to achieve optimal performance in whatever scenario is thrown at it. If manual overclocking is the way to do it, then I'll manually overclock. If XFR can provide that, then I'll use XFR.

 

[sm625]

Here are a bunch of benchmarks. These are taken directly from passmark PerformanceTest 9.0. Note: Actual clocks are not listed on these graphs. The true clocks are as follows:

The good news is that turbo for the Ryzen is apparently disabled.

Without a source, I am not going to allow these "benchmarks"
Please add a source.
Markfw
Anandtech Moderator

Wow

As the post clearly says:

These are taken directly from passmark PerformanceTest 9.0.

These are taken directly from passmark PerformanceTest 9.0.

These are taken directly from passmark PerformanceTest 9.0.

These are taken directly from passmark PerformanceTest 9.0.

These are taken directly from passmark PerformanceTest 9.0.

These are taken directly from passmark PerformanceTest 9.0.

How much clearer can I possibly make it?

Just... Wow.

Saying you did a test and got certain results using software X is not the same as linking to a recognized web site that details everything about the test. I could use my E5-2683, do the tests, and claim something else, or even photoshop the results.
Call me out again, and its another infraction
Markfw
Anandtech Moderator

 

[The Stilt]

My brain has the consistency of extremely overboiled cabbage today..Please humour me.

So you're saying that if you Purchase a Ryzen cpu listed as 3.6/4.0ghz, then that is 4.0 MACF?

No.
3.6GHz MACF and 4.0GHz MSCF.

 

[OrangeKhrush]

Based on comparing apples to oranges? Sigh, you really like these comparisons..

And what more can be done? yesterday I posted how CPUZ and Passmark detected no Boost when I turned it off in BIOS, today I posted how AIDA, Passmark are able to detect a standard Boost and manipulated boost.

Why is no boost profile showing on any of the AMD parts database? The likely reason is it was disabled in BIOS.

Another indication is the benches it did well at followed the Benches used in Handbrake, Blender and the gaming leaks which shows Ryzen at low frequency scoring high, this was also shown up in Passmarks Integer, Floating Point and Singlethread hardly different to what has been seen with AMD using a 3.4Ghz base clock which puts in in that Haswell to slightly above range clock for clock, this was not a far fetched expectation, just one that everyone with some kind of inuendo has tried to kill anticipation with.

You can believe that or invisible turbos but I think something is happening and this is far from a Bulldozer release and we haven't even got one X370 with final sample bench yet, just some gimped runs.

 

[The Stilt]

And what more can be done? yesterday I posted how CPUZ and Passmark detected no Boost when I turned it off in BIOS, today I posted how AIDA, Passmark are able to detect a standard Boost and manipulated boost.

Why is no boost profile showing on any of the AMD parts database? The likely reason is it was disabled in BIOS.

Another indication is the benches it did well at followed the Benches used in Handbrake, Blender and the gaming leaks which shows Ryzen at low frequency scoring high, this was also shown up in Passmarks Integer, Floating Point and Singlethread hardly different to what has been seen with AMD using a 3.4Ghz base clock which puts in in that Haswell to slightly above range clock for clock, this was not a far fetched expectation, just one that everyone with some kind of inuendo has tried to kill anticipation with.

You can believe that or invisible turbos but I think something is happening and this is far from a Bulldozer release and we haven't even got one X370 with final sample bench yet, just some gimped runs.

AMD's TSC has been running at P0 (highest non boosted, i.e base state) frequency since Bulldozer (still applies to Zen). Unless you overclock (i.e modify the P0 PState for higher frequency) the OS and all of the other programs which do not constantly monitor the frequency by using other counters (e.g. HPET) detect the base clock speed. Look at Cinebench for example. For AMD FX-8350 or FX-8370 parts it detects the CPU speed as 4.0GHz (base) regardless if turbo is enabled or not.

 

[Erenhardt]

I'm wondering if XFR offers something manual OC can't really properly provide, at least not with any reasonable convinience.
For example, if all cores can turbo up to 4.5, but a single core can do 4.8. You can always go into the BIOS and disable the rest of the cores to reach that 4.8, and then switch back to 4.5 with all cores when you're done with the task, but that would be incredibly annoying for everyday use.

If XFR can detect load on a specific core and boost it to the edge, while powering down the rest of the cores to a level that allows that single core to reach 4.8GHz, then it does something that would be straight up annoying to attempt in manual overclocking.

Basically, if XFR can optimize itself for specific application loads, then I'm getting a SKU with one. As fun as overclocking is, I'd like for my computer to be able to achieve optimal performance in whatever scenario is thrown at it. If manual overclocking is the way to do it, then I'll manually overclock. If XFR can provide that, then I'll use XFR.

Exactly.

I sorta tried that with PD, but ultimately, I failed.
I defined turbo to 4.5Ghz on half cores and 5.0GHz on single core as well as 4.4 base clock. Ofc that required voltage adjustement. +0.15 voltage offset seems to work nice without turbo enabled, but with turbo, it boosted already high turbo voltage above 1.5Volts, which was not something I'm comfortable.

Lowering offset voltage to 0.1V turned out to be a disaster. While a 5.0GHz single core turbo was nice and stable at 1.5V, the base clock was not. It was running 4.4 below 1.3V, and was unstable. After a lot of tweaking, trial and error, it was apparent that I would need to go with my base clock below 4.0GHz or live with a crazy single core boost above 1.5V. Both of which sucked. So now, Im running flat 4.5 without turbo at all...

If XFR can do that fine tuning automagically and actually adjust itself properly (something I couldn't do manually) then it is simply amazing for everybody.

 

[OrangeKhrush]

I am not refering to how it reports the CPU ID on Cinebench or any benchmark for that matter

This Bench is run at 4.6ghz but Cinebench doesn't report information straight from the registry like CPU, AIDA, Passmark do.

What I am asking or more the point I am getting at with the screenshots is that Passmark detects turbo when enabled and reflects the turbo or overclock max, if it is disabled it shows N/A I did this by disabling a 4460's turbo from BIOS and it showed up as N/A and in the CPU information screen where there is normall a 3200mhz [3400mhz] the bracket was omitted like the AMD CPU. The only real inferance is that it was disabled from BIOS like the 4460's as Passmark can detect max boost range and report it even if the CPU is in C STATE

 

[Dresdenboy]

No.
3.6GHz MACF and 4.0GHz MSCF.

So do they advertise MACF as "base"? Being "max" - this could mean slightly lower clocks for heavy loads (e.g. Prime95 w/ SMT).

 

[lolfail9001]

Uhhh, Mr Mod, Passmark does detail about everything about the test except actual clock. Also, does not look like he was the one who did the test in the first place.

 

[The Stilt]

So do they advertise MACF as "base"? Being "max" - this could mean slightly lower clocks for heavy loads (e.g. Prime95 w/ SMT).

I'm not entirely sure since some models have XFR while others don't.
Base clock is still base clock, regardless. If a SKU has 3.4GHz base clock and no XFR, then that is the MACF at the same time.

 

[Agent-47]

4.0 GHz should be maximum single core frequency, maximum all core frequency is probably not going to be advertised (that's how it is on FX/APUs).

my FX 6300 us advertised as 3.5/3.9Ghz. but it turbos to 3.9Ghz on all core and 4.1 on single core.

EDIT: sorry, apparently it is advertised as a 4.1 Ghz. i stand corrected

No.
3.6GHz MACF and 4.0GHz MSCF.

i hope you are not correct on this one

 

[bjt2]

XFR doesnt really interest me. It bothered me with my 960 on the GPU side of the spectrum.

What I want, is pstate control, just like Wattman gives me for AMD gpus or just like good ol' phenommsrtweaker or Stilt's software (a shame that it wasn't further developed).

XFR is better than that type of OC. Is a continuos self calibration of p-states, based on temperature and load. No other OC mean can beat it, at same TDP.

 

[bjt2]

Exactly.

I sorta tried that with PD, but ultimately, I failed.
I defined turbo to 4.5Ghz on half cores and 5.0GHz on single core as well as 4.4 base clock. Ofc that required voltage adjustement. +0.15 voltage offset seems to work nice without turbo enabled, but with turbo, it boosted already high turbo voltage above 1.5Volts, which was not something I'm comfortable.

Lowering offset voltage to 0.1V turned out to be a disaster. While a 5.0GHz single core turbo was nice and stable at 1.5V, the base clock was not. It was running 4.4 below 1.3V, and was unstable. After a lot of tweaking, trial and error, it was apparent that I would need to go with my base clock below 4.0GHz or live with a crazy single core boost above 1.5V. Both of which sucked. So now, Im running flat 4.5 without turbo at all...

If XFR can do that fine tuning automagically and actually adjust itself properly (something I couldn't do manually) then it is simply amazing for everybody.

Exactly: it monitors actual temperature 8that depends also of previous loads, ambient and obviously cooler) to clock to the max supported clock in THAT moment.

 

[KTE]

^Sounds good on paper, remains to be seen how it performs in reality.

Imagine reviews in deep winter compared to summer!

Sent from HTC 10
(Opinions are own)

 

[CatMerc]

^Sounds good on paper, remains to be seen how it performs in reality.

Imagine reviews in deep winter compared to summer!

Sent from HTC 10
(Opinions are own)

Yup, the devil is in the details. The concept sounds incredible, the execution however will decide how it will pan out.

 

[OrangeKhrush]

I'm not entirely sure since some models have XFR while others don't.
Base clock is still base clock, regardless. If a SKU has 3.4GHz base clock and no XFR, then that is the MACF at the same time.

Spoiler: Image

If it cannot detect turbo, either it is off or that motherboard cannot detect it, in which case it is hard to claim it was boosting

 

[lolfail9001]

Imagine reviews in deep winter compared to summer!

I can almost imagine XFR refusing to hit advertised boost clocks at summer.

If it cannot detect turbo, either it is off or that motherboard cannot detect it

I'll reveal you a terrible secret: it is not motherboard that reports clocks.

 

[OrangeKhrush]

I can almost imagine XFR refusing to hit advertised boost clocks at summer.

I'll reveal you a terrible secret: it is not motherboard that reports clocks.

The OS will detect it, but it doesn't and there is a complete lack of proof that it was working

 

[wembley]

It's a party https://twitter.com/VideoCardz/status/830756074126970880 he he he

Sent from my HTC One_M8 using Tapatalk

 

[AtenRa]

If it can reach 4.5GHz and it has higher IPC than IvyBridge at those prices, then ill get a 6-Core RYZEN with a mini-iTX X300 mobo to replace my Core i7 3770K @ 4.4GHz.

 

[inf64]

The train is rolling and brakes are broken . Chu-chu !!!

 

[Shivansps]

If it can reach 4.5GHz and it has higher IPC than IvyBridge at those prices, then ill get a 6-Core RYZEN with a mini-iTX X300 mobo to replace my Core i7 3770K @ 4.4GHz.

Well acording to passmark

If its running at 3.4Ghz its exactly Haswell IPC, at 3.8Ghz it may be below Haswell IPC, personally i think its running at 3.4Ghz, the table says "Turbo Clock: N/A" like with the I3s, it may also be because the program is unable to detect it,